[PPT] - Automated Reasoning for Situational Awareness Peter Baumgartner, PowerPoint Presentation

SLIDE 1

www.data61.csiro.au

Automated Reasoning for Situational Awareness

Peter Baumgartner, Alexander Krumpholz

Supply Chain Integrity Digital Mission

SLIDE 2

Situational Awareness - Systems of Interest

Factory Floor
Are the operations carried out according to the schedule?
Food Supply Chain
Are goods delivered within 3 hours and stored below 25℃?
Why is the truck late?
Where did the strawberries/honey come from?
Data Cleansing
Does the database have complete, correct, accurate and relevant data?

SLIDE 3

Situational Awareness - Systems of Interest

Factory Floor
Are the operations carried out according to the schedule?
Food Supply Chain
Are goods delivered within 3 hours and stored below 25℃?
Why is the truck late?
Where did the strawberries/honey come from?
Data Cleansing
Does the database have complete, correct, accurate and relevant data?

Build a system that automatically derives such analysis?

SLIDE 4

Situational Awareness - Systems of Interest

Factory Floor
Are the operations carried out according to the schedule?
Food Supply Chain
Are goods delivered within 3 hours and stored below 25℃?
Why is the truck late?
Where did the strawberries/honey come from?
Data Cleansing
Does the database have complete, correct, accurate and relevant data?

Build a system that automatically derives such analysis?

SLIDE 5

Automated Reasoning for Situational Awareness

Why is the truck late? Are the tomatoes still fresh?

?

GPS coordinates Temp sensor Paperwork Log DB

Events…

SLIDE 6

Automated Reasoning for Situational Awareness

Why is the truck late? Are the tomatoes still fresh?

?

GPS coordinates Temp sensor Paperwork Log DB

Incomplete/noisy/erroneous data Need domain knowledge (“fresh”?)

Why this is hard

Events…

SLIDE 7

Automated Reasoning for Situational Awareness

Why is the truck late? Are the tomatoes still fresh?

?

GPS coordinates Temp sensor Paperwork Log DB

Incomplete/noisy/erroneous data Need domain knowledge (“fresh”?)

Why this is hard

+ +

Events…

SLIDE 8

Automated Reasoning for Situational Awareness

Why is the truck late? Are the tomatoes still fresh?

?

GPS coordinates Temp sensor Paperwork Log DB

Incomplete/noisy/erroneous data Need domain knowledge (“fresh”?)

Why this is hard

+ +

Our logic-based approach

Domain modelling (first principles)

Stuck at warehouse / fresh Traffic jam / not fresh OR

What-if reasoning and explanations Events…

SLIDE 9

Automated Reasoning for Situational Awareness

Why is the truck late? Are the tomatoes still fresh?

?

GPS coordinates Temp sensor Paperwork Log DB

Incomplete/noisy/erroneous data Need domain knowledge (“fresh”?)

Why this is hard

+ +

Our logic-based approach

Domain modelling (first principles)

Stuck at warehouse / fresh Traffic jam / not fresh OR

What-if reasoning and explanations Events…

Implemented in the Fusemate system

SLIDE 10

Demo

T

Observation: truck is in Sydney at the warehouse

SLIDE 11

Demo

T

Observation: truck is in Sydney at the warehouse

SLIDE 12

Demo

T

Observation: tomatoes are loaded

SLIDE 13

Demo

T

Observation: tomatoes are loaded

SLIDE 14

Demo

T

Assumption as per schedule: truck is on the road

SLIDE 15

Demo

T T+1

Assumption as per schedule: truck is on the road

SLIDE 16

Demo

T T+1

Report: truck is on the road

SLIDE 17

Demo

T T+1 T+2

Report: truck is on the road

SLIDE 18

Demo

T T+1 T+2

Conclusion: truck is on the road for too long - tomatoes are no longer fresh

SLIDE 19

Demo

T T+1 T+2

Conclusion: truck is on the road for too long - tomatoes are no longer fresh

SLIDE 20

Demo

T T+1 T+2

Report: actually, at T+1 truck was still in Sydney warehouse

SLIDE 21

Demo

T T+1 T+2

Report: actually, at T+1 truck was still in Sydney warehouse

SLIDE 22

Demo

T T+1 T+2

Conclusion: tomatoes are still fresh at T+2

SLIDE 23

Demo

T T+1 T+2

Conclusion: tomatoes are still fresh at T+2

SLIDE 24

Demo

T T+1 T+2

?

T+3

No information at T+3

SLIDE 25

Demo

T T+1 T+2 T+3

T+3: What if truck is on the road?

SLIDE 26

Demo

T T+1 T+2 T+3

T+3: What if truck is on the road?

SLIDE 27

Demo

T

OR

T+1 T+2 T+3

T+3: What if truck is on the road? At Canberra warehouse?

SLIDE 28

Demo

T T+1 T+2 T+3

Report: truck at Canberra warehouse

SLIDE 29

Demo

T T+1 T+2 T+3

Report: truck at Canberra warehouse

SLIDE 30

Fusemate in more detail…

SLIDE 31

Fusemate Model = If-Then Rules

T1 T2

Time

SLIDE 32

Fusemate Model = If-Then Rules

T1 T2 T1 < T ≤ T2

Time

NO!

SLIDE 33

Fusemate Model = If-Then Rules

T1 T2 T1 < T ≤ T2

NO!

Time

✔

SLIDE 34

Fusemate Model = If-Then Rules

If T1 T2 T1 < T ≤ T2 and then

NO!

Time

SLIDE 35

Fusemate Model = If-Then Rules

If T1 T2 T1 < T ≤ T2 and then

NO!

If in(T1, B, C) & T1 < T2 NOT exists T s.th. T1 < T ≤ T2 & unload(T, B, C) then in(T2, B, C)

Time

SLIDE 36

Fusemate Model = If-Then Rules

If T1 T2 T1 < T ≤ T2 and then

NO!

If in(T1, B, C) & T1 < T2 NOT exists T s.th. T1 < T ≤ T2 & unload(T, B, C) then in(T2, B, C)

default reasoning: “not known” (≠ “known not”)

Time

SLIDE 37

Fusemate Model = If-Then Rules

If T1 T2 T1 < T ≤ T2 and then

NO!

If in(T1, B, C) & T1 < T2 NOT exists T s.th. T1 < T ≤ T2 & unload(T, B, C) then in(T2, B, C)

default reasoning: “not known” (≠ “known not”)

Time

If at T1 and at T2 then revise: or revise:

🠩
🠩

SLIDE 38

Fusemate Model = If-Then Rules

If T1 T2 T1 < T ≤ T2 and then

NO!

If in(T1, B, C) & T1 < T2 NOT exists T s.th. T1 < T ≤ T2 & unload(T, B, C) then in(T2, B, C)

A model is a set of if-then rules plus …

default reasoning: “not known” (≠ “known not”)

Time

If at T1 and at T2 then revise: or revise:

🠩
🠩

SLIDE 39

Model = If-Then Rules + Scala Class Hierarchy

abstract class Item { val perishable: Boolean } abstract class Fruit extends Item { val perishable = true } abstract class Vehicle { val speed: Int } case class Truck(id: Int, time: DateTime, load: Set[Item]) extends Vehicle with LogicFact { val speed = 80 val rules = List( Truck(id, t, load - item) :- Unload(id, t, item), … ) def hasPerishableLoad = load contains { _.perishable = true } }

S c a l a c

d

e

SLIDE 40

Model = If-Then Rules + Scala Class Hierarchy

abstract class Item { val perishable: Boolean } abstract class Fruit extends Item { val perishable = true } abstract class Vehicle { val speed: Int } case class Truck(id: Int, time: DateTime, load: Set[Item]) extends Vehicle with LogicFact { val speed = 80 val rules = List( Truck(id, t, load - item) :- Unload(id, t, item), … ) def hasPerishableLoad = load contains { _.perishable = true } }

S c a l a c

d

e If-then rule extension

SLIDE 41

Model = If-Then Rules + Scala Class Hierarchy

case class Truck(id: Int, time: DateTime, load: Set[Item]) … { … @rules(id, time, load) val rules = List( Fail :- Unload(id, time, item), IF (! load contains items) Truck(id, next, load - item) :- Step(next, time), Unload(id, time, item), NOT(t < time, Unload(id, t, item) ) ) }

SLIDE 42

Model = If-Then Rules + Scala Class Hierarchy

case class Truck(id: Int, time: DateTime, load: Set[Item]) … { … @rules(id, time, load) val rules = List( Fail :- Unload(id, time, item), IF (! load contains items) Truck(id, next, load - item) :- Step(next, time), Unload(id, time, item), NOT(t < time, Unload(id, t, item) ) ) }

Scala library “set”

SLIDE 43

Model = If-Then Rules + Scala Class Hierarchy

case class Truck(id: Int, time: DateTime, load: Set[Item]) … { … @rules(id, time, load) val rules = List( Fail :- Unload(id, time, item), IF (! load contains items) Truck(id, next, load - item) :- Step(next, time), Unload(id, time, item), NOT(t < time, Unload(id, t, item) ) ) }

Macro annotation Scala library “set”

SLIDE 44

Model = If-Then Rules + Scala Class Hierarchy

case class Truck(id: Int, time: DateTime, load: Set[Item]) … { … @rules(id, time, load) val rules = List( Fail :- Unload(id, time, item), IF (! load contains items) Truck(id, next, load - item) :- Step(next, time), Unload(id, time, item), NOT(t < time, Unload(id, t, item) ) ) }

Macro annotation

case (Step(next, time1), Unload(id1, time2, item)) if time1 == time && time2 == time && id1 == id ! (model exists { case Unload(id2, t, item1) if id2 == id &&   t < time && item1 == item => true case _ => false } }) => Truck(id, next, load - item)

Expansion Scala library “set”

SLIDE 45

Modelling Paradigm Summary

Fix erroneous event data and revise explanations Derive multiple plausible explanations

Inferences on basis of incomplete information

👎

D e c l a r a t i v e

Rich data structures Java connectivity Inference engine

Rules Scala

👎 Infrastruture

SLIDE 46

System Architecture Explana'ons Inference Engine

Fusemate

Domain Model If-then rules

Updat

EPCIS Events

REST

Q/A/C

Domain Model - e.g., If item I is unpacked from a container C at 3me T then I must have been packed into C at some 3me S < T Explana'ons - The inference engine derives a set of plausible models consistent with the EPCIS events so far EPCIS Events - EPCIS events are sent to fusemate as they become available Update - Plausible models are updated on every new EPCIS event and command provided by user Q/A/C - Where was item I at 3me T? Item I was unpacked from container C at 3me T and loc L. Item I has never arrived at L!

SLIDE 47

Case Study

SLIDE 48

Case Study - Deer Supply Chain

2013

Gary Hartley New Zealand RFID Pathfinder Group January 2013

The Use of EPC RFID Standards for Livestock and Meat Traceability

SLIDE 49

Case Study - Deer Supply Chain

2013

Gary Hartley New Zealand RFID Pathfinder Group January 2013

The Use of EPC RFID Standards for Livestock and Meat Traceability

12 events - from farm (NZ) to retailer (DE) encoded in EPCIS

SLIDE 50

Case Study - Deer Supply Chain

2013

Gary Hartley New Zealand RFID Pathfinder Group January 2013

The Use of EPC RFID Standards for Livestock and Meat Traceability

7

Process Step 4 - Animals arrive at Mountain River Processors’ stun box Figure 5.7 - Stun Box Figure 5.8 - RFID reader at Stun Box Figure 5.7 illustrates animals in the location of the stun box. Note the RFID ear tags in the ears of the

animals. Figure 5.8 illustrates the RFID antenna setup at the stun box.

Process Step 5 - Cartons of finished Venison cuts packed into cartons at Mountain River processor and moved from the boning room into chiller room Figure 5.9, Figure 5.10 and Figure 5.11 illustrate the affixing of EPC UHF RFID tags on the cartons in the boning room and moving of cartons of finished venison cuts into the chiller room in preparation for loading the shipping container. Figure 5.9 - UHF RFID tags used on cartons Figure 5.10 - UHF RFID tags positioned on cartons Figure 5.11 – Tagged cartons moving from boning room to chiller room

12 events - from farm (NZ) to retailer (DE) encoded in EPCIS

SLIDE 51

Case Study - Deer Supply Chain

2013

Gary Hartley New Zealand RFID Pathfinder Group January 2013

The Use of EPC RFID Standards for Livestock and Meat Traceability

7

Process Step 4 - Animals arrive at Mountain River Processors’ stun box Figure 5.7 - Stun Box Figure 5.8 - RFID reader at Stun Box Figure 5.7 illustrates animals in the location of the stun box. Note the RFID ear tags in the ears of the

animals. Figure 5.8 illustrates the RFID antenna setup at the stun box.

Process Step 5 - Cartons of finished Venison cuts packed into cartons at Mountain River processor and moved from the boning room into chiller room Figure 5.9, Figure 5.10 and Figure 5.11 illustrate the affixing of EPC UHF RFID tags on the cartons in the boning room and moving of cartons of finished venison cuts into the chiller room in preparation for loading the shipping container. Figure 5.9 - UHF RFID tags used on cartons Figure 5.10 - UHF RFID tags positioned on cartons Figure 5.11 – Tagged cartons moving from boning room to chiller room

Table 6.3 - Commissioning event - tagging of animals ‘ ’.

12 events - from farm (NZ) to retailer (DE) encoded in EPCIS

SLIDE 52

Case Study - Deer Supply Chain

2013

Gary Hartley New Zealand RFID Pathfinder Group January 2013

The Use of EPC RFID Standards for Livestock and Meat Traceability

7

Process Step 4 - Animals arrive at Mountain River Processors’ stun box Figure 5.7 - Stun Box Figure 5.8 - RFID reader at Stun Box Figure 5.7 illustrates animals in the location of the stun box. Note the RFID ear tags in the ears of the

animals. Figure 5.8 illustrates the RFID antenna setup at the stun box.

Process Step 5 - Cartons of finished Venison cuts packed into cartons at Mountain River processor and moved from the boning room into chiller room Figure 5.9, Figure 5.10 and Figure 5.11 illustrate the affixing of EPC UHF RFID tags on the cartons in the boning room and moving of cartons of finished venison cuts into the chiller room in preparation for loading the shipping container. Figure 5.9 - UHF RFID tags used on cartons Figure 5.10 - UHF RFID tags positioned on cartons Figure 5.11 – Tagged cartons moving from boning room to chiller room

Table 6.3 - Commissioning event - tagging of animals ‘ ’. –